Real-time 3D shape sensing based on GPU-accelerated OFDR with dynamic sweep-range compensation

In this paper, we proposed and demonstrated a dynamic sweep-range compensation method, namely fixed-range k-domain resampling (FR-KDR), for a real-time 3D shape sensing system based on multicore fiber (MCF) and optical frequency domain reflectometry (OFDR). The compensation performance under zero-strain condition was investigated, and the strain root mean square error (RMSE) was reduced from 13.98 µε to 1.27 µε. Moreover, both 2D and 3D shapes were successfully reconstructed, and the position error along the sensing MCF was further analyzed. The maximum position deviations of the reconstructed 2D and 3D shapes with an MCF length of 700 mm were 13.49 mm and 16.95 mm, respectively. Ultimately, the FR-KDR method and the wavelength shift demodulation algorithm were deployed on the graphics processing unit (GPU) to perform parallel computation, and the dynamic stability of the real-time OFDR system was evaluated. The time consumption under different zero-padding sizes and sweep ranges was analyzed, and an update rate of 23.8 Hz was achieved at a spatial resolution of 5 mm with a sweep range of 20 nm and a zero-padding size of 8192. Moreover, the comparison between CPU serial computation and GPU parallel computation confirmed that GPU acceleration significantly improved the data-processing efficiency.